Architecture and isolation
CenterChat uses managed hosting, database, authentication, and AI providers. Workspace records are scoped by authenticated ownership checks, server-side routes, and relational access rules. Public widgets use signed sessions, allowed-origin controls, and rate limits.
Access and administration
Administrative access is allowlisted by role. Sensitive admin changes create audit records. Production credentials stay in server-side environment storage and logs redact common secret fields. Customers remain responsible for strong account passwords, access reviews, and prompt removal of departed operators.
Data protection and minimization
Transport encryption is provided by the hosting and service platforms. Conversation retention is configurable and scheduled deletion removes linked messages, leads, handoffs, and action records. Operational analytics accepts only allowlisted metadata and excludes conversation content.
Application safeguards
The service applies request validation, same-origin protection for authenticated mutations, SSRF checks for outbound calls, signed webhooks, atomic quota claims, provider fallbacks, prompt-injection detection, failure logging, and restricted widget origins. No safeguard removes every risk; consequential workflows require human review.
Operations and recovery
Health checks, structured logs, scheduled failure summaries, encrypted database backups, dependency updates, automated tests, and deployment quality gates support operations. Recovery procedures and backup restoration should be tested regularly by the operator.
Report a security concern
Send a concise report with affected URL, impact, reproduction steps, and contact details. Do not access data that is not yours, degrade service, use automated high-volume testing, or publicly disclose an unresolved issue. We will acknowledge and prioritize credible reports according to risk.
Questions about this policy?
Ask the team that runs CenterChat.